Kedro node execution explanation

[datajoely]

kedro-streaming/src/kedro_streaming/io/spark_streaming_dataset.py

Line 304 in 1144426

def _load(self) -> DataFrame:

Could you explain how this load mechanism knows when to cede execution to the rest of the pipeline. In many of the streaming applications allow you deploy queries/views which act as 'running applications' that are always running.

Kedro has ended up being a very batch centric methodology - how do we reconcile this concept with streaming?